WO2005121809A2

WO2005121809A2 - Method and computer programme for determining operating parameters in a roller bearing and a roller bearing which may be analysed thus

Info

Publication number: WO2005121809A2
Application number: PCT/EP2005/006143
Authority: WO
Inventors: Alfred Pecher
Original assignee: Schaeffler Kg
Priority date: 2004-06-08
Filing date: 2005-06-08
Publication date: 2005-12-22
Also published as: JP2008501959A; EP1754066A2; DE102004027800B4; US7716018B2; US20080091361A1; WO2005121809A3; DE102004027800A1

Abstract

The invention relates to a method and a computer programme for the determination of operating parameters, such as particularly, the rotational speed, rotational direction, the radial force, the axial force or structural noise events of a rolling roller bearing (20), to which a sensor arrangement (26) is fixed, delivering a sinusoidal signal, dependent on the rotational position of the bearing on the rotation of the roller bearing (20), which is sampled at sampling points k, k+1, k+2, . For the determination of the rotational speed, the rotational direction, the radial force and/or the axial force, a mean value is determined by means of estimation, and the operating parameters determined on the basis of a corrected signal which gives no offset by subtraction of the estimated mean value from the sensor signal. For determination of the structural noise events, the sensor signals are filtered in a high-pass filter (59) and structural noise occurring is determined by means of estimation of a statistical moment for at least two orders. The invention further relates to a roller bearing (20), provided with an analytical device (50), for carrying out said method or the computer programme.

Description

METHOD AND COMPUTER PROGRAM FOR DETERMINING OPERATING PARAMETERS IN A BEARING BEARING AND A READABLE BEARING BEARING THEREFORE

The present invention relates to a method and a computer program product (also called computer program or software for short) according to claims 1 and 19 for determining operating parameters in a rolling bearing and a rolling bearing which can be evaluated with the aid of the aforementioned method according to claim 20.

Rolling bearings are used in every machine in the industrial area. Due to the constantly increasing demands on the service life and the operational safety of such machines, there is an increasing need for individual operating parameters such as the speed, the direction of rotation, forces acting on the bearing in the axial and radial direction, and disturbances in the bearing, which are generally referred to as structure-borne noise events, to be recorded in order to derive, among other things, indications of potentially critical situations for the warehouse and / or the machine.

In principle, there are a few possibilities to find the necessary procedural automatism for the determination of speed and force. For example, the local extrema could be found by differentiating the curve and searching for the zeros of the first derivative. The times for the occurrence of an extremum can be read from this, from which the functional value of the original curve at the corresponding point in time is then obtained. However, it must be expected that the sensor signal is overlaid with interferers, so that the approach of differentiation does not lead to a usable result. In addition, values of the signal curve would have to be buffered using this method. Such problems are often also solved by detecting the local maxima and minima of the curve via the time shift of a window with a width adapted to the signal frequency. However, this procedure is only practical if the frequency remains constant or changes only slowly in order to be able to readjust the window width accordingly. When used in the warehouse, the speed changes are too fast, so that such an approach must be excluded from the outset.

Spectral evaluation methods for determining the signal frequency and the speed to be deduced therefrom are also unsuitable for the application, since the real-time character is lost through Fourier transformation or Fast Fourier transformation and the temporal averaging contained therein, since fluctuations in the frequency are no longer detected can. In addition, a considerable number of function values of the curve must be buffered here in order to obtain the necessary one to achieve spectral resolution.

There is therefore the difficulty that the methods available for calculating these operating parameters, owing to the large amounts of data, the storage capacity required for these amounts of data and the necessary computing power, can only be carried out in external computers, i.e. not on site in a chip in the rolling bearing itself. There is also the problem that the large number of data lines, which are necessary for transporting the recorded rolling bearing data to the outside to the external computer, are difficult to accommodate in a small rolling bearing.

The problem of data lines can be solved by integrating sensors and evaluation electronics in the form of several sensors with assigned microcomputers in the vicinity of the rolling bearing, if possible even in the rolling bearing itself. with a so-called IQ Bearing - solve. The demands placed on development are particularly high, since this local integration of sensors and evaluation electronics is to be carried out without changing the external geometric shape of the bearing. This naturally increases the demands on the microelectronic components due to the increased environmental influences of temperature, lubricants and coolants and the small size of the required hardware components. In particular, due to the small size of the microprocessors, the functional integration due to the necessary procedure poses a particular challenge, since due to the limited computing capacity for the evaluation of the data supplied by the sensor, only a limited tool is available and nevertheless a reliable determination of the operating parameters of the bearing is required.

The above considerations result in the fact that storage in the planned application has to be largely dispensed with and accordingly the development of the method has to be adapted to the geometric conditions, not only with regard to the simplest possible arithmetic operations, but in particular also to the number of for the intermediate storage of samples necessary gates.

The present invention is based on the object of creating the most efficient but safe method for determining operating parameters in a rolling bearing, which is intended to offer the possibility of using an electronic circuit which is located in the vicinity of the rolling bearing and preferably in the rolling bearing itself to be carried out, the method also being intended to be implementable as a computer program. Furthermore, a rolling bearing is to be created which can be or is connected to an evaluation device in which the operating parameters are determined efficiently and reliably.

This object is achieved with a method according to claim 1 or with a computer program solved according to claim 19 and a rolling bearing according to claim 20. Advantageous developments of the invention are the subject of the dependent claims.

The inventive method according to claim 1 and the inventive computer program according to claim 19 can be integrated into an electronic circuit or a chip even in the rolling bearing itself in a very small space, reacts "intelligently" to dynamic changes in the rolling bearing and is capable of operating parameters with high To determine accuracy.

The rolling bearing according to the invention according to claim 20 comprises an evaluation device with which the method according to claims 1 to 18 or alternatively the and / or storable on a storage medium (for example RAM, ROM, CD, DVD, floppy disk, hard disk, flash memory, etc.) or can be executed via a network accessible computer program according to claim 19. It is taken into account here that the evaluation device, which is preferably designed as an ASIC in a chip, has only limited computing capacities due to its placement in the groove of the outer ring. The width of the chip is determined a priori by the width of the outer ring of the bearing. In addition, the chip is not too long in the circumferential direction of the outer ring, since otherwise the chip located in the groove would be disproportionately bent due to the curvature of the outer ring, and a defect would therefore result. Due to the limited chip area, the mathematical or signal-theoretical possibilities that can be realized by the module are indeed very limited, but are completely sufficient for carrying out the method according to the invention. In particular, no evaluation methods are used that rely on the intermediate storage of a large amount of signal values.

Further advantages, features and special features of the invention result from the following description of preferred, but not restrictive, embodiments of the invention on the basis of the schematic and not to scale drawings. Show it:

1 is a schematic representation of a rolling bearing according to the invention,

2 schematically shows the arrangement of strain gauges in a Wheatstone bridge circuit in the external groove of a rolling bearing according to the invention,

3 shows a typical example of a sensor signal generated in the roller bearing according to the invention,

4 shows a sensor signal in which a structure-borne noise event occurs,

5 shows a block diagram of an ASIC according to the invention, 6 shows a sensor signal in which reference points for evaluating the operating parameters are given,

7 shows a representation similar to FIG. 2 for an arrangement for detecting the axial force acting on the bearing,

8 is an illustration for explaining the measuring principle for the detection of the direction of rotation,

9 shows a graph of the time course of the estimate of the mean value,

10 shows a graph of the transient response of the sensor signal freed from the mean value,

11 shows the frequency response of the IIR high-pass filter in the structure-borne sound branch,

12 shows a sensor signal for a bearing under radial load and two structure-borne noise events that have occurred,

13 is a graph of the structure-borne noise signal at the output of the IIR high-pass filter,

14 shows a graph of the output signal of the IIR high-pass filter, the signal on the left having structure-borne noise, while on the right it is free of structure-borne noise,

FIG. 15 shows a graph of the same section of the signal as in FIG. 14, but the usual second-order statistical moment of the signal was calculated recursively,

16a to 16h are flow diagrams which illustrate a preferred embodiment of the method according to the invention for determining all operating parameters in all details.

A rolling bearing according to the invention and the theoretical foundations for a method for determining the operating parameters are first described using a preferred embodiment, and then this method according to the invention for determining these operating parameters is described in detail using flow diagrams.

1 and 2 represent the main components of the roller bearing 20 according to the invention, a so-called intelligent bearing. The representation is symbolic and serves to illustrate, but is not to be regarded as restrictive.

The intelligent bearing is intended to provide the user with information about the force acting on the roller bearing 20 (hereinafter also abbreviated to "bearing"), about the speed and the direction of rotation of the bearing. Furthermore, a possibility is to be created for the detection of rolling bearing faults and of impacts transmitted to the rolling bearing 20, which are noticeable as a structure-borne sound signal. Such a roller bearing 20 comprises an inner ring 21 and an outer ring 22, on the outside of which a circumferential longitudinal groove 23 is provided. Rolling elements 24 are arranged between the inner ring 21 and in the outer ring 22, so that the inner ring 21 is rotatable relative to the outer ring 22. For the recording of the data, sensor arrangements 26 - also called "sensors" 26 for short - are used, which in the preferred embodiment are four strain gauges 31 to 34 combined to form a Wheatstone bridge circuit, which are accommodated in an external longitudinal groove 23 in the outer ring 22 and the like Resistance changes due to the rolling over by the rolling elements 24. Furthermore, a circuit board 28 arranged in the groove 23 is shown schematically, which produce the conductor connections between the individual strain gauges 31 to 34 of each sensor 26 and the conductor connections between the sensors 26 and the evaluation devices 50 described later. The direction of movement of the rolling elements 24 is indicated by an arrow A in FIG. 2.

The groove 23 and thus the printed circuit board 28 is located in the entire circumference of the outer ring 22 with sensors 26 (strain gauges 31 to 34) arranged at equidistant intervals and a corresponding evaluation device 50 for each sensor, as will be explained in more detail later. The resulting sensor signal 40 is to be evaluated in a suitable manner by the evaluation device 50, preferably electrical circuits in the form of user-specific circuits, so-called ASICs (Application Specific Integrated Circuits). The method according to the invention, which runs in the ASICs, for determining the operating parameters sought must be dimensioned in such a way that consistent online evaluation of the data is possible despite the restricted installation space and thus the limited chip size. The complete unit of sensor and intelligent evaluation hardware thus represents a so-called "smart sensor", which makes it possible to provide the potential customer with extensive information about the bearing 20 and its operating parameters in real time.

For the configuration of the method according to the invention, the geometric arrangement of the sensors 26 in the outer ring 22 must be taken into account, since this determines the course of the signal 40 supplied by the sensor 26 and thus also specifies the procedure for the signal-theoretical treatment of the sensor signal 40. The four strain gauges 31 to 34 of a sensor 26 connected to form a Wheatstone bridge are arranged in the longitudinal direction of the groove 23 in such a way that their distance from one another corresponds to half the distance between the rolling elements (cf. FIG. 2). This ensures that two strain gauges 31 and 33 or 32 and 34 of a bridge are always rolled over by rolling elements 24 at the same time. The first strain gauge 31 of the subsequent Wheatstone bridge in the groove 23 is also again half the distance between the rolling elements and the last strain gauge 34 of the previous bridge. This arrangement results in ^z / ₂ sensors 26 in the bearing for z rolling elements. This ensures that Force influences can be measured from all radial directions. Alternatively, fewer or even more than ^z / ₂ sensors can be present, for example ^z / ₂ -1, ^z / ₂ +1, z + 1, or 2 z sensors. Generally speaking, the strain gauges (hereinafter also abbreviated as strain gauges) are preferably connected to form bridges such that the distance between the first strain gauge and the third strain gauge and the distance between the second strain gauge and the fourth strain gauge are an integral multiple of the rolling element distance. It is also possible to connect only two strain gauges (for example the first and second strain gauges) instead of a third or fourth strain gage, each with a non-strain-sensitive resistor to form a Wheatstone half bridge. Instead of the DMS mentioned, other sensors such as piezoelectric or magnetic sensors can also be used. In this case, a different ratio of sensors to rolling elements can result.

2 schematically shows the geometric distribution of strain gauges and rolling elements 24 in the groove 23 of the outer ring 22 for a sensor. When DMS1 31 and DMS3 33 are rolled over, the bridge output voltage increases with a constant supply voltage U _B with increasing deformation of the strain gauges caused by the rolling bearing pressure U _A (t) on. The strain gauges connected crosswise in the bridge circuit are rolled over or deformed synchronously, which increases the increase in the output voltage. In the further course, the strain gauges DMS1 31 and DMS3 33 are increasingly relieved until they have reached their original resistance value and the bridge is almost balanced. When the strain gauges DMS2 32 and DMS4 34 are then rolled over, the same profile of the output voltage occurs, only because of the dimensioning of the bridge circuit with the opposite sign, so that the approximately sinusoidal profile of FIG. 3 ultimately results. In dashed lines, an alternative embodiment of a sensor 29 with four strain gauges 36, 37, 38 and 39 is given, each of which has an approximately triangular shape and thereby enables the detection of an axial force acting on the bearing 20. Of course, there are other ways of designing the strain gauges for measuring the axial force.

The greater the force acting on the bearing 20 or the corresponding sensor 26, the greater the surface pressure exerted by the rolling element 24 on the strain gauges. This leads to a change in the resistance values of the bridge circuit and thus to a correspondingly high deflection of the bridge output voltage. Accordingly, there is a deterministic relationship between the force exerted on the bearing and the output voltage U _A (t) of the bridge circuit, which can be used to evaluate the elongation or the force caused by the elongation. According to FIG. 3, there is an average signal swing of approximately 0.03V.

Furthermore, the frequency f of the signal 40 corresponds to the rollover frequency of the rolling elements 24 on the fixed outer ring 22, which is mathematical about the rotational speed of the shaft in the following Connection stands:

The variable z indicates the number of rolling elements 24, D _w is the diameter of the rolling elements, D _{pw is} the pitch circle diameter and α is the contact angle, which is to be set to zero in the cylindrical roller bearing. With n, the speed of the shaft is denoted in ^υ / _Mm , which has to be converted to the unit ^U / _{Hz in} order to obtain the frequency in Hz.

As can also be seen from FIG. 3, the sensor signal 40 has an offset or offset which differs from sensor to sensor and which arises due to the process when the thin-film structures of a bridge are applied to the roller bearing steel with resistance values which are not exactly the same. The unbalanced distribution of the resistance values results in the DC component at the bridge output. A further offset can arise during operation of the rolling bearing due to changes in the ambient conditions, for example due to changes in temperature.

In principle, a surface defect in the rolling bearing or a structure-borne noise event caused by the rollover of dirt particles is manifested in the application by a signal component in the sensor signal that is higher in frequency than the approximately sinusoidal fundamental vibration. 4 shows an example of such a rolling element fault.

5 shows a block diagram for an evaluation device 50 according to the invention for determining the operating parameters of a roller bearing 20, this evaluation device preferably being in the form of an ASIC which can be arranged in the groove 23 of the outer ring 22 of the bearing 20. Accordingly, after the digitization of the sensor signal 40 in an A / D converter 56 with subsequent filtering of high-frequency interference components by a low-pass filter 57, preferably an FIR (Finite Impulse Response) filter, the further processing of the sensor signal is carried out in two modules arranged in parallel. This FIR filter 57 is preferably sixteenth order at a cutoff frequency of 12 kHz. For example, it could also have a higher order, but then requires more space on the ASIC 50. At the beginning of each individual module, an evaluation of the frequency range is carried out, which contains the relevant information for the evaluation in the branch. Low pass filtering takes place in the upper module, preferably in a further FIR filter 58 likewise of the sixteenth order, with a cutoff frequency of f _g = 3 kHz, since the signal frequencies to be expected for the planned applications can reach a maximum of the cutoff frequency of the filter 58. When dimensioning the filter limit, more emphasis was placed on the suppression of higher-frequency signal components than on the amplitude-appropriate transmission by the filter module. Any errors that may occur due to the amplitude damping are later compensated for by increasing the calculated signal swings. The desired operating parameters of the bearing are then determined in the form of the speed, the radial force (corresponding to the Stroke of the signal 40), the axial force and the direction of rotation, which is represented schematically by corresponding function blocks 51 to 54. In the lower module, the frequency correlated with the rotational speed is suppressed by a high-pass filter 59, preferably an MR (Infinite Impulse Response) high-pass filter of the sixth order, and in function block 55 the generally higher-frequency component of the structure-borne noise signal is weighted and then suitable for post-processing evaluated. As can be seen from the flowcharts of FIGS. 16a to 16h, the determination of the individual operating parameters does not necessarily take place separately in individual function blocks (also called partial methods or "evaluation units") 51 to 55, but preferably in one, according to the schematic representation in FIG. 2 only method in which the individual sub-methods or evaluation units (function blocks 51 to 55) are interlocked and lead to results in real time.

The filter limits of the FIR filters are preferably selected so that any bearing designs can be evaluated with the same ASIC. The filter limits therefore represent a compromise between the achievable signal-to-noise ratio SNR in the upper branch of signal evaluation and the cost distribution for the intelligent warehouse over many bearing types and designs. In the lower branch of the ASIC, in this embodiment there was a need to implement a sixth order programmable IIR filter. However, this also makes it possible for a bearing type to adjust the filter in rough frequency steps at varying speeds and thus to carry out the best possible analysis of structure-borne noise events. The features of the sensor signal determined in the blocks "direction of rotation", "stroke", "speed", "axial force" and "structure-borne noise" are obtained via an SPI (serial peripheral interface) located in the groove 23 of the outer ring 22, but not shown here. -Bus forwarded to a DSP (digital signal processor), which is also not shown, located outside the warehouse. The DSP is responsible for organizing the data exchange in the warehouse, programming the ASICs in the warehouse or querying the data from the warehouse. In addition, the DSP can calculate a total force for the force acting on the individual sensors 26, which results from the respective signal strokes and the geometric distribution of the measuring points. However, it essentially forms the interface between the intelligent warehouse and the customer application, which fetches data from the warehouse with the standardized interface provided by the DSP and acts on the system according to the specifications or monitors the system.

There is now a need to find a method which is as efficient as possible for evaluating the sensor signal 40 and thus for determining the parameters of speed, direction of rotation, axial force and stroke. One possibility for the evaluation with simple computing means arises if the sensor signal is crossed with an imaginary auxiliary line, as shown in FIG. 6.

The parameters of the curve of the sensor signal 40 which are necessary for an evaluation and are related to the auxiliary line are all entered in FIG. 6. This results, for example, from the two _Times t _aUf i and t _auf2 , which mark the crossing points between an auxiliary line 42 and the successive rising edges of the curve, the period of the sensor signal and thus the shaft speed according to equation (1). The speed can thus be determined online within a very short time. The intersection points of the falling flank of the sensor signal 40 with the auxiliary line 42, t _{a 1} and t _ab2 , can of course also be used for the calculation of the rotational speed. The given sensor geometry also makes it possible to determine the speed in a half-wave, which seems particularly useful at low speeds. In the combination of the evaluation of the rising and falling edge intersections, the access time to the value of the shaft speed is halved again, since a new value is available for each half period, so that the information is quickly available even at slow speeds. Once the crossing points between the auxiliary line 42 and the sensor signal 40 have been recognized, it can be determined very easily whether the evaluation is currently on the ascending or descending curve. If the evaluation is being carried out for the ascending curve, the method according to the invention searches for the maximum value in the sample values of the sensor signal 40 which are present over time until the point of intersection between the falling edge and the auxiliary line 42 is reached. The smallest value in the pending samples is then searched for until the next crossing point for the rising edge is detected. This results in the maximum y _max and the minimum y _min for each signal _period , which subtracts the signal _swing and thus represents the counterpart to the radial force recorded for this sensor 26. Here, too, the access time to the stroke value and thus to the radial force can be halved by continuously alternating the differences ym _a xi-ymim and y _max2 - y _mιn ι etc.

By slightly modifying the strain gauge arrangement within the measuring bridge shown in FIG. 7, the auxiliary line concept can also be used to install a criterion for the direction of rotation in the signal curve. For this purpose, a pair of strain gauges (in the picture DMS2 32 'and DMS4 34') is arranged in the sensor 27 with respect to the strain gauges 32 and 34 shown in dashed lines for clarity in accordance with the arrangement known from FIG be overrun at an earlier time. As a result, an asymmetry can be generated in the course of the sensor signal 40, which can be used to detect the direction of rotation. 8 shows the measuring principle on a positive half-wave of the sensor signal. The relationships in the illustration are exaggerated in comparison to the real asymmetry in order to clarify the principle. Due to the displacement of the strain gauges 32 'and 34', the extrema of the signal 40 shift relative to one another in the time direction, so that, for example, the maximum, depending on the direction of rotation, either to the left or to the right of the time center (with "0" on the abscissa in FIG. 8 designated) from t _auf1 and t _{a 1} , whereby t _aUf i and t _ab1 do not change in the arrangement according to FIG. 8 compared to the arrangement according to FIG. 2. The direction of rotation results from the sign of the expression, for example

~ 'max i - \ -> As can be easily recognized, the evaluation of the sensor signal 40 can be carried out very easily in this way. However, there is again the difficulty here in designing the auxiliary line 42 in a suitable manner. In principle, the mean value of the offset signal 40 is useful as an auxiliary line 42. However, according to the conventional calculation of the mean value, this would also require the corresponding storage space for the temporary storage of the values for at least one period. If, for example, a slow speed of 100% _in. Is assumed, a signal frequency of approx. 11 Hz results. With the sampling frequency of the system of 78 kHz used, approx. 7000 values with a word width of 14 bits each have to be buffered for a period, which - according to an integration density based on a structure size of 0.35 μm - corresponds to a chip area of approx. 90 mm ² . Because of the geometrical specifications, such a method would not be practicable in the case of conventional averaging; furthermore, online evaluation would not be possible.

A solution can, however, be found by recursively / adaptively estimating the desired mean value of the signal 40 as the aforementioned auxiliary line 42 intersecting the signal 40. This recursive / adaptive estimate of the mean represents a preferred embodiment of the present invention; it is generally sufficient if an auxiliary line is selected which intersects the signal 40 and is essentially parallel to its envelope or essentially follows this signal.

The following scalar equation (3) results for the task formulated above for estimating the mean value of the curve shape provided by the strain-sensitive sensors 26: + 1) = E {γ} (k) + K (k) ■ (y (k + 1) - E {Y} (k)) (3)

The stochastic variable E {Y} (k) represents the linear expected value of the sensor signal y (k) at the time k and K (k) represents the feedback gain or in other words the adaptation constant. This is an essential hurdle for the evaluation of the Taken operating parameters of the bearing 20 and the simple procedure briefly outlined above for determining the speed, direction of rotation, radial force and axial force can be carried out. The mean value calculated in this way has a signal-adaptive character and follows the offset, which may change slowly over time due to temperature influences. Proper evaluation of the sensor signal 40 is thus always guaranteed, even if the signal 40 has an unsteady character with regard to its mean value.

For practical use it can be advantageous to calculate the feedback gain K (k) dynamically depending on the noise processes. The design of the adaptive estimate with regard to the speed and consistency of the estimate can, for example, be based on stochastic criteria. The result is a feedback gain K (k) that changes over time. At the beginning of the adaptation, the values of K (k) are very high, which leads to a rapid settling to the desired value. In the further course of time the _ _

Gain exponentially. This significantly reduces the influence of the disturbances on the estimate. However, the functionality that arises as a result, with maximum efficiency, is out of proportion to the effort required to calculate the exponential curve. For this reason, the noise influences are not considered further in the method according to the invention. Rather, a static gain K (later also called cO) is of greater practical use, because overall the calculation effort for estimating the mean value remains very low. All in all, according to equation (3) one needs a register for the static factor K or c and for the mean value E {Y} (k). The computing effort for the determination of the estimated values for the mean is based on a few fundamental arithmetic operations.

FIG. 9 shows the course of the estimate of the mean value 43 for the sensor signal 40 specified in FIG. 3 according to the method discussed.

For the calculation of the operating parameters, the mean value obtained for each sampling cycle is subtracted from the current sensor signal value.

The feedback gain K = 0.0001 was selected for the display. 10 shows the settling behavior of the sensor signal 45 freed from the mean value 43. The mean value is reached after about 0.5 seconds, so that the parameters can be queried at the latest from this point in time. However, from t = 0.05 s the sensor signal 45 already crosses with the zero line, which represents the auxiliary line 42. From this point in time, the method according to the invention delivers result values, although the values themselves are still slightly faulty, since the crossing points change relative to one another at a constant speed.

In the next step, the theoretical basics for a suitable method for the evaluation of structure-borne noise events, i.e. the detection of structure-borne noise events, are described.

The frequency response of the sixth-order IIR filter 59 in the structure-borne branch of FIG. 5 can be programmed in its frequency characteristic by downloading coefficients via the SPI bus. For the detection of rolling bearing defects, a high-pass characteristic is imposed on the IIR filter in order to separate the high-frequency structure-borne noise signals from the approximately sinusoidal basic signal according to FIG. 3.

11 shows a typical application-related frequency response of the filter 59. The cut-off frequency for the pass band is 1 kHz. Attenuation is attenuated at 60 dB, in the passband attenuation is negligibly small at 0.0002 dB. With the aid of the high-pass filter 59, the structure-borne noise event, as shown in the circled section of FIG. 4 or can be seen in FIG. 12, is extracted in a first step. The result of the filtering is shown in FIG. 13. In principle, the filter delivers a very good result due to its good properties in the pass and blocking range by separating the structure-borne noise event from the low frequencies exactly in real time, so that the fundamental signal of the sensor signal 40 can no longer be seen in the filtered signal 44. Based on this, automated detection of a structure-borne noise event could be designed in such a way that a threshold is drawn parallel to the time axis. As soon as the threshold of the output signal 44 of the IIR filter 59 is exceeded, there is a structure-borne noise event. However, the selectivity at the output of the filter 59 between the structure-borne noise event and the periods in which no structure-borne noise occurs is not very high. If the threshold is low enough, it is in principle possible that it is constantly exceeded due to the noise quantities superimposed by the filter signal 44. On the other hand, to avoid these false alarms, the threshold must not be too high, since otherwise not only the noise levels will not reach the threshold, but also less pronounced structure-borne noise events can no longer be detected. For this reason, it is very difficult to find the optimal threshold function, especially if no automation can be used due to the lack of computing capacity.

One possibility of improvement is to increase the ratio between structure-borne noise signal and noise floor of the filter output and thus the signal-to-noise ratio in this branch. For this purpose, the calculation of the ordinary second-order statistical moment at the output of filter 59 should preferably be used:

Based on the quadratic evaluation of the output values of the IIR filter 59

Structure-borne noise events are emphasized much more strongly from the noise processes, and it is much easier to set a threshold for reliable detection, as was shown by tests on real data. However, here too there is the restriction not to use any methods with a high storage space requirement, so that the determination of the usual second statistical moment in the conventional form of equation (4) cannot be carried out using N values. Based on the previously described method for determining the speed, an alternative method for the calculation can also be found here. Analogous to equation (4) for estimating the linear expected value, the recursive signal-adaptive approach for the usual second statistical moment is obtained: E {Y ² Yk + \) = E {Y ² k) + K (k) - (y ² ( k + \) - E {Y ² ) (*)) (5)

The accuracy of the detection of structure-borne noise can be further improved by linking the results of the recursive equations accordingly and evaluating the resulting signal. For example, this results in the recursive signal-adaptive version of the variance calculation: σ _y ² (k + \) = E {γ ² k + \) - [E {γ} (k + l) J (6)

The relevance of the calculated estimated values to the operating parameters which can be determined using the method according to the invention will now be briefly discussed below. When determining the operating parameters "speed", "direction of rotation", "radial force" and "axial force", the result of the mean value estimation no longer appears, but this does not detract from the importance for the quality. Rather, this forms the basis for the "lean", ie. H. efficient evaluation of the data and for the determination of the desired operating parameters.

When determining the direction of rotation, an average value of approximately -0.13 ms was obtained for the deviation of the maximum from the central axis of the positive half-wave of the sensor signal, which corresponds to approximately ten samples at the sampling frequency of f _a = 78 kHz. The sign during the entire recording was always less than zero, so that a direction of rotation can be clearly assigned. The other direction would accordingly produce a positive sign in the deviation.

In contrast to the recursive calculation of the linear expected value for the aforementioned operating parameters, the determination of the structure-borne noise events is directly included in the result value. In order to illustrate the mode of operation of the method according to the invention, the output signal 46 of the IIR filter 59 shown in FIG. 14 is used, only the front area of the output signal 46 containing a structure-borne noise signal. The rear area is taken up by a roller bearing without structure-borne noise signal in order to make a direct comparison.

If the recursive adaptive estimate of the usual statistical moment of the second order according to equation (5) is applied to this signal 46, the result of the resulting signal 48 is a clear highlighting of the first section, as shown in FIG. 15. The occurrence of a structure-borne noise event can thus be detected much better via a threshold value than was previously the case with the signal 46 according to FIG. 14.

The evaluability of the signal 48 obtained can be further increased if the aforementioned recursive adaptive estimate of the usual second-order moment according to equation (5) is applied to the signal 48 and then the variance according to equation (6) is calculated using the equation (3) becomes.

If, on the other hand, the variance is already calculated for signal 46, this also brings about an improvement in the evaluability of signal 46, but not to the extent that is the case when calculating the variance of signal 48. However, this saves some calculations, which results in quick processing. In addition to the clearer signal-to-noise ratio in the structure-borne sound branch of FIG. 5, the time information can be evaluated when the result is exceeded via the threshold and the error source can be assigned on the basis of the period resulting therefrom. The feedback gain was set to K | = 0.8 in this example.

The accuracy of the method according to the invention is relevant in two directions: the resolution in the amplitude range and the resolution in the time range. The precision of the values for the signal swing and structure-borne noise determined by the intelligent bearing depend on the quantization and modulation of the analog-digital conversion. The speed accuracy is predetermined by the time resolution determined by the scanning grid of the system, as is the direction of rotation, the latter measure representing a qualitative - not quantitative - characteristic of the bearing.

Depending on the voltage supply of the ASIC implemented for the tests, which according to the specification may be a minimum of 3.0 V and a maximum of 3.6 V, and the so-called dynamic signal amplitude for the analog-digital converter from -3 mV to +3 mV per volt supply voltage, with an assumed average supply voltage of 3.3 V, the voltage range for the sensor signal to be converted is between -9.9 mV and +9.9 mV. This corresponds to a maximum signal swing of 19.8 mV. The word length of the implemented analog-digital converter is 14 bits in the fault-free case, so that a voltage value of 1.2 μV is represented on a digit. The maximum load of the cylindrical roller bearing considered here is assumed to be 50 kN. Distributed over the entire value range with a maximum modulation of the analog-digital converter, this results in a theoretical accuracy of approx. 3 N per digit. The requirements for the method according to the invention are thus more than met, since the transmission system of the surrounding construction, the bearing and the sensor results in a substantially coarser quantization. Measurements gave a resolution in the range of 200 N. The quantization in the time direction is predetermined by the sampling frequency of the system, which is exactly 78.125 kHz in the system used for the tests. The theoretically maximum error of the speed - and the direction of rotation - is derived from the reciprocal of the sampling frequency: .T _a = 1 / f _a = 12.8 μs. With a maximum possible rollover frequency of all possible bearings of 3 kHz, this results in a maximum error of 3.8%. This represents a very good compromise between the relative error in the measurement and the data rate of the system.

A method is described below with reference to the flow diagrams shown in FIGS. 16a to 16h as a preferred embodiment, with which all of the above-mentioned operating parameters can be determined in parts of the method which are interlinked with one another. For the sake of clarity, however, the description is subdivided into the determination of a single operating parameter, the parameter "T" being omitted for the elapsed time in order to simplify it; the elapsed time is therefore implicit in the parameter "k". The connection of the flowcharts among themselves is indicated by the connection points V1 to V8.

As already stated, each sensor 26 delivers a wave-shaped signal 40, which is dependent on its rotational position, when the roller bearing 20 rotates, for example as shown in FIG. The signal 40 is sampled at - preferably equidistant - sampling points k, k + 1, k + 2, ..., and the resulting sampling values x ^* (k), x * (k + 1), x ^* (k + 2) , ... are further processed in accordance with the respective parts of the method according to the invention.

For k = 0, the signal 40 is first sampled to determine its instantaneous value as the sample start value x ^* (0) and a start value for the estimated value E (x * (k = 0)) is defined as the expected value of the signal 40. Then the run variable k is increased by 1. The signal 40 is then sampled at the time k + 1 to determine its instantaneous value as a (k + 1) th sample value x * (k + 1).

rotational speed

16a, a (k + 1) -th estimated value E (x ^* (k + 1)) as the sum of the previous, k-th estimated value E (x ^* (k)) and the product of an adaptation constant cO is shown in FIG. 16a with the result of subtracting this kth estimate E (x ^* (k)) from the (k + 1) th sample x * (k + 1). Then, in step S101, the (k + 1) -th estimated value E (x ^* (k + 1)) calculated in step S100 is subtracted from the k + 1-th sample value x ^* (k + 1), and the difference becomes a (k + 1) th calculation value x (k + 1) is used for the further method. The following applies to cO: 0 <cO <1. The last term in curly brackets in the calculation of step S101 is not relevant for the calculation of the rotational speed and can therefore be disregarded. In step S102, the sign of this calculated (k + 1) th calculation value x (k + 1) is compared with the sign of the corresponding kth calculation value x (k) from the previous run through this method part. The process then branches into two branches:

If the signs compared in step S102 are the same, ie the signal 40 has not undergone a change of sign and, in other words, has not undergone a zero crossing, k is increased by 1 and method steps S100 to S102 are repeated. In other words, none of the method steps S302 to S316 according to FIG. 16c is necessary for determining the speed. If the signs compared in step S102 are not the same, ie the signal 40 has made a zero crossing, a further case distinction is compared in step S203 according to FIG. 16b as to whether the (k + 1) th calculation value x (k + 1) from current step S101 is greater than the corresponding k-th calculated value x (k), which was already determined when step S101 was run through previously. In other words, it is determined whether the signal 40 is on a rising edge or a falling edge. Method steps S108, S110 and S112 according to FIG. 16a are also not necessary for determining the speed. Incidentally, the determination of the zero crossings offers the additional advantage that the method always knows in which phase of the sampling of the signal 40 one is currently. The zero crossings set as it were "Navigation system" and facilitate the assignment of the measured values recorded. It should be mentioned here that the zero crossings, for example when the shape of the sensors 26, 27 or 29 changes, are "lashed down" on the x-axis, as it were, while the shape of the signal changes, as is the case, for example, due to the asymmetry in FIG. 8 is shown.

If it is found in step S203 with x (k + 1)> x (k) that the signal is on a rising edge, it is checked in step S210 whether the current calculated value x (k + 1) is exactly zero. If this is the case, the point in time at which this calculated value occurred is fixed in step S214 as the point in time (cf. FIG. 6) of the zero crossing on the rising edge. If this is not the case, however, the time _{t2 on} the zero crossing on the rising edge is determined in step S212 by linear interpolation between the two calculated values x (k + 1) and x (k). This linear interpolation represents a preferred procedure of the method according to the invention. Alternatively, as a useful approximation - in order to save the calculation for the linear interpolation - the point in time immediately before the occurrence of the zero crossing or the point in time immediately after its occurrence could be used, in the further method then the same point in time before or after the occurrence of the zero crossing would always have to be taken into account. In step S404 of FIG. 16d then becomes the "ascending" period length .DELTA.T _to by forming the difference between two zero crossings on the rising edge, namely the just determined "new" time and the determined in the previous cycle corresponding "old" time t _at i , calculated. Then, in step S406, the "new" time t _{auf2 is stored} as the "old" time t _aut i, so that it can be used when a "rising" period length ΔT _auf is _recalculated . Step S408 is not required for the determination of the speed, and in step S410 an output flag for the output of the "ascending" period length ΔT _is set to - ie, a command is given to the processor to calculate it - and this is output as a measure of the speed , It is clear that when the method is run for the very first time, ie when the "new" time t _aUf2 is first determined, no "old" time t _aιlf ι is available. A "rising" period length ΔT _aU f is _{therefore output} for the first time when a "new" time t _{auf2 has been} calculated for the second time. Then k is increased by 1 and the method continues with step S 100.

On the other hand, if x (k + 1)> x (k) does not result in step S203, this means that the signal is on a falling edge. In this case, steps S230, S232 or S234 are carried out analogously to steps S210, S212 or S214 described above, the time t _{a 2 of} the zero crossing being determined on the falling edge instead of t _aUf2 . Analogous to steps S404 and S406, the "falling" period length ΔT _ab is then calculated in steps S504 and S506 according to FIG. 16e by forming the difference between the "new" time t _{a 2} and the corresponding "old" time ta ₁ and output in step S510 as a measure of the speed. Thereafter, k is also increased by 1 and the method continues with step S100. This determines the speed - twice per period: in the form of the "rising" period length ΔT _up and the "falling" period length ΔT _down .

In principle, it would also be possible to determine the period length by the distance between two minima or two maxima, since these minima or maxima form a relatively broad "crest" and are therefore not as sharp as the zero crossings, their use is particularly useful for Noise less precise and therefore not preferred.

When determining further operating parameters, steps which have already been described are not shown again in detail in order to avoid repetitions, especially since they do not need to be carried out again anyway.

axial force

In determining the axial force, steps S100 and S102 are the same as in determining the speed, and the only difference in step S101 to step S101 used in determining the speed is that in determining the calculated value x (k + 1 ) an auxiliary variable hub hlf is added, d. that is, the last term in curly brackets in the calculation of step S101 is taken into account this time. This means nothing else than that the auxiliary line 42 should not be identical to the zero line, but rather that it lies below. The auxiliary variable hub-hlf can be chosen freely, but it is preferred that it is composed of a product of a number (hlf) between 0 and 0.5 with the signal stroke (hub).

Steps S203 and steps S210, S212 and S214 for the rising edge and steps S230, S232 and S234 for the falling edge are carried out in exactly the same way as in the determination of the rotational speed.

For the rising edge, the time interval (Δt_nhw), how long the negative half-wave lasts, is then determined in step S402 by subtracting the "new" time t _{ab2 of} the falling edge from the "new" time t _{auf2 of} the rising edge becomes. The illustration in FIG. 6 deviates from this since it does not take into account that the corresponding values t _au n. T _{aU 2} . ta _b i, "continue to hike" in different process runs, ie that their designation changes successively as the process proceeds. Thereafter, step S406 is executed, which represents such a successive change.

In parallel to the execution of steps from S101 to S406, a further method branch is carried out, which is referred to as "parallel branch" in FIGS. 16a and 16f and begins with step S601 in FIG. 16f. Step S601 differs from step S101 only in that the auxiliary variable hub hlf is not added, but is subtracted. That means nothing else than that the intersection of the signal 40 should take place with a corresponding auxiliary line lying further above. For better differentiability for the calculated value x (k + 1) from step S101, the calculated value resulting in step S601 is designated y (k + 1), this distinction being retained in steps S602, S603, S610, S612, S614, S630, S632 and S634 , which otherwise run identically to steps S102, S203, S210, S212, S214, S230, S232 and S234 in the "main branch". In order to clarify that the zero crossings determined in the "main branch" differ from those determined in the parallel branch, the latter are marked with a " ^* ". For the falling edge, the time interval (Δt_phw), how long the positive half-wave lasts, is then determined in step S638, namely by the fact that the "new" time t * _{on2 of} the rising edge from the "new" time t ^* _{ab2 of} the falling _edge Edge is subtracted. Then step S506 analogous to step S406 is carried out.

In step S408 and / or step S508, the time interval Δt_phw determined in step S638 can then be subtracted from the time interval Δt_nhw determined in step S402, the sign of this difference Δt_phw - Δt_nhw indicating the direction ax_dir of the axial force acting on the bearing 20 and in step S410 and / or S510 can be output. In addition, in step S409 and / or S509, the amount ax_for of the difference Δt_phw - Δt_nhw can be determined as a measure of the magnitude of the axial force and output in step S410 and / or S510.

This also determines the axial force acting on the bearing 20.

radial force

When determining the radial force, steps S100, S101 and S102 are the same as when determining the speed. If the signs compared in step S102 are the same, ie if no zero crossing has taken place, it is checked in step S302 whether the sign of the (k + 1) th calculation value x (k + 1) of step S101 is identical to 1, ie , the signal 40 is in the positive half wave. If so, it is checked in step S304 whether this (k + 1) th calculation value x (k + 1) is greater than the previously determined maximum max_value of the signal 40, and if so, this (k +1) th calculated value x (k + 1) is stored as the new maximum max_value of the signal 40, then k is increased by 1 and the method is continued with step S100. If this is not the case, the stroke (stroke) of the signal 40 is determined in step S308 by forming the difference max_wert-min_wert between its maximum max_wert and its minimum min_wert, the determination of the minimum min_wert being described later. Of course, in the method according to the invention, the stroke can only be calculated if both the maximum max_value and the minimum min / vert are present. Then k is increased by 1 and the method continues with step S100. If it is found in step S302 that the sign of the (k + 1) th calculation value x (k + 1) is not identical 1, ie the signal 40 is in the negative half-wave, in an analogous manner in steps S314, S315 or S318 determines or stores the minimum min_value of signal 40 and calculates the stroke. Then k is increased again by 1 and the method continues with step S100. It should be mentioned that the signal swing can also be calculated and output by the inverse difference, i.e. minj / vert - max_wert, and thus also twice per signal period. As soon as it has emerged in step S102 that a sign change or a zero crossing has taken place, an output flag is set in step S108 that the stroke can be output as a measure of the radial force acting on the bearing 20. In step S110, it is checked whether the stroke is greater than a previously determined maximum signal stroke hubjnax. If this is the case, the currently determined stroke is stored as the new maximum signal stroke hub_max. This serves to detect strong impacts. The value for the maximum signal lift hubjnax can be deleted by the user at any time.

Then again in step S203, a case distinction is made as to whether the signal is on a rising edge or on a falling edge. In the former case, the maximum max_value, in the second case the minimum min_value of the signal 40 is reset to zero, k is increased by 1 and the method is continued with step S100.

The radial force is thus also determined as a further operating parameter of the roller bearing 20.

It can be seen from the above description that a minimum or maximum is only accepted as such if a zero crossing between these two extremes has been detected. This ensures that in the event of a "dent" in the signal, no sub-extreme associated with it is recorded as a (false) minimum or maximum.

Drehrichtunq

When determining the direction of rotation, steps S100, S101 and S102 are the same as in

Determination of the speed.

If the signs compared in step S102 are the same, i. that is, if no zero crossing has taken place, it is checked in step S302 (FIG. 16c) whether the sign of the (k + 1) th calculation value x (k + 1) of step S101 is identical to 1, i. that is, signal 40 is in the positive half wave. If so, it is checked in step S304 whether this (k + 1) th calculation value x (k + 1) is greater than the previously determined maximum max_value of the signal 40, and if so, the time max_wert_t becomes in step S306 (corresponds to i in FIG. 8 and the expression (2)) for which this (k + 1) th calculation value x (k + 1) (ie max_value) occurred, then k increased by 1 and the method with step S100 continued. If it appears in step S302 that the sign of the (k + 1) th calculation value x (k + 1) is not identical 1, i. that is, the signal 40 is in the negative half-wave, it is checked in an analogous manner in step S314 whether this (k + 1) th calculation value x (k + 1) is smaller than the previously determined minimum value of the signal 40, and if this is the case, the time min_value_t at which this (k + 1) th calculation value x (k + 1) (ie min_value) occurred is stored in step S315, then k is increased by 1 and the method is continued with step S100 ,

If the signs compared in step S102 are not the same, ie the signal 40 is one 16b in a further case distinction is compared in step S203 according to FIG. 16b, whether the (k + 1) th calculation value x (k + 1) from the current step S101 is greater than the corresponding kth calculation value x (k) which was already determined when step S101 was run through previously. In other words, it is determined again whether the signal 40 is on a rising edge or a falling edge. If the signal 40 is on the rising edge, the direction of rotation is determined in step S204 in the following way: A minimum auxiliary variable min_t_ref, the meaning of which will be explained later, is subtracted from the time min_wert_t of the occurrence of the minimum of the signal 40 and the result is obtained 2 divided, and from this the result of subtracting the minimum auxiliary variable min_t_ref from the time max_wert_t of the occurrence of the maximum of the signal 40 is subtracted. The sign of this value is used as an indication of the direction of rotation sym of the rolling bearing 20, after which an output flag for the direction of rotation sym is set in step S208 and this is output. Before the output, however, the time min_wert_t of the occurrence of the minimum of the signal 40 is stored in step S206 as a new minimum auxiliary variable min_t_ref. This also clarifies the meaning of these minimum auxiliary variables min_t_ref. Then k is increased by 1 and the method continues with step S100. However, if the signal 40 is on the falling edge, the direction of rotation sym is determined in step S224 (similarly to step S204) in the following manner: A maximum auxiliary variable max_t_ref, the meaning of which is also explained later, is determined from the time max_wert_t of the occurrence of the Maximum of the signal 40 subtracted and the result divided by 2, and from this the result of subtracting the maximum auxiliary variable max_t_ref from the time min_wert_t of the occurrence of the minimum of the signal 40 is subtracted. The sign sign of this value is used again as an indication of the direction of rotation sym of the roller bearing 20, after which an output flag for the direction of rotation sym is set in step S228 and this is output. Before the output, however, the time max_wert_t of the occurrence of the maximum of the signal 40 is stored in step S216 as a new maximum auxiliary variable max_t_ref. The meaning of these minimum auxiliary variables max_t_ref is also clarified.

The direction of rotation of the roller bearing 20 is thus also determined.

In the above, the use of a uniform adaptation constant cO in the respective parts of the method for determining the speed, the direction of rotation, the axial force and the radial force implicitly assumed that this adaptation constant is the same everywhere in the calculation of the expected value of the signal 40. However, it is clear that, if necessary, different adaptation constants can also be used in the different parts of the process, if this brings advantages for the calculation of the different operating parameters.

The description of the determination of structure-borne noise events now follows.

Kάrperschall For this purpose, for k = 0, the signal 40 is first sampled to determine its instantaneous value as the sampling start value x ^* (0). An estimated value E (x ^* (k = 0)) is not required as the expected value of the signal 40, but the second usual statistical moment E {V} (k = 0) of the filter sample value v (k = 0), as will be explained later , Then the run variable k is increased by 1. The signal 40 is then sampled at the time k + 1 to determine its instantaneous value as a (k + 1) th sample value x ^* (k + 1), etc.

The samples x ^* (k), x ^* (k + 1), x ^* (k + 2), ..., abbreviated x ^* (k) in FIG. 16g, preferably pass through an IIR high-pass filter 59 in step S702 at least 5th order, whose filter coefficients are freely programmable. The low-frequency component of the sensor signal 40 is thereby filtered out. The filter samples v (k), v (k + 1), ..., abbreviated here to v (k), are then present at the output of the filter 59, which is shown in FIG. 13. In step S704, the second ordinary statistical moment E {V ² } (k + 1) is calculated based on equation (5) at the “new” filter sample value v (k + 1). For this purpose, the second ordinary statistical moment E {V ² } (k) from the square v ² (k + 1) of the "new" filter sample v (k + 1) is subtracted from the "old" filter sample v (k), this difference with a The adaptation constant d is multiplied, and the result of the multiplication is then added to the "ordinary" filter sample v (k) at the second ordinary statistical moment E {V ² } (k). The following applies to d: 0 <d <1. This ordinary second-order statistical moment is shown on the left in FIG. After that, the output signal Output can be output as it is in step S710 as a quantitative error signal, in addition the output signal Output can be compared in step S718 with a threshold value and - if it exceeds this - can be output as a qualitative value, which in step S720 is referred to as "structure-borne sound output flag". In addition, in step S712 it can be compared whether the output signal Output exceeds a previously recorded (user-erasable) maximum output signal Output_max, and in step S714 this can then be stored as a new maximum output signal Output_max, after which it can be output in step S716, what is called "Output Flag Output_max".

For further processing, the output values E {V ² } (k + 1) from the calculation of step S704 are abbreviated u (k) for reasons of clarity. Then in step S706, the calculation of the second ordinary statistical moment known from step S704 is carried out, but based on the values u (k), in order to convert the second ordinary statistical moments E {U ² } (k + 1) to u (k +) 1) to calculate. Instead of the adaptation constant d used in step S704, an adaptation constant c2 is used, for which the following applies: 0 <c2 <1. However, it can also apply: c1 = c2. In step S707, the expected value E {U} (k + 1) of u (k + 1) is calculated based on equation (3): For this purpose, the expected value E {U} (k) of is calculated from u (k + 1) u (k) subtracted, this difference multiplied by an adaptation constant c3 (for which the following applies: 0 <c3 <1), and the result of the multiplication is then added to the expected value E {U} (k). c3 need not necessarily be different from d and / or c2. Then in step S708, the variance σ ² (k + 1) is calculated by squaring the result of step S707 from the result of step S706 is subtracted. Thereafter, an output (optionally in steps S710 and / or S716 and / or S720) can be carried out with the abovementioned possibilities, with the steps S706 to S708 being more accurate in detecting and differentiating the sensor signal and noise than in one Output already after step S704.

The structure-borne noise occurring in the roller bearing 20 is thus also recorded.

Many studies have shown that the method according to the invention very reliably determines the desired operating parameters and works optimally for the task at hand. By using the equations for the linear expected value and the usual statistical moment of the second order, it is possible to determine the operating parameters safely and quickly by evaluating the data in the ASIC in a very small space. The procedure is u. a. it is also very efficient because it mainly uses comparisons and additions and no complex multiplications or divisions, and is based on simple storage with a minimal space requirement.

In the above description, it was assumed that the method according to the invention is carried out in an ASIC, that is to say is carried out in a hard-wired arrangement. However, it is clear that the method according to the invention can also be implemented in the form of software, i. H. a computer program that can be executed in a normal computer.

It should be noted that the features of the invention described with reference to individual embodiments, such as, for example, individual steps from the flowcharts or the description of the theoretical principles of the method, may also be present in other embodiments, unless stated otherwise or from technical reasons Prohibited by itself.

Claims

CLAIMS:

1. A method for determining operating parameters, such as in particular the speed, the direction of rotation, the radial force, the axial force or structure-borne noise events, of a rotating roller bearing (20), to which a sensor arrangement (26; 27; 29) is attached, which a Rotation of the rolling bearing (20) delivers a wave-shaped signal (40) which is dependent on its rotational position and is sampled at sampling points k, k + 1, k + 2, ..., whereby sampling values x ^* (k), x ^* (k + 1 ), x * (k + 2), ..., with an auxiliary line (42) that intersects the signal (40) for the determination A) of the speed, the direction of rotation, the radial force and / or the axial force and is essentially parallel to its envelope, is constructed and the operating parameters are determined on this basis, and / or B) of structure-borne noise events, the sampling values x * (k), x * (k + 1), x * (k + 2) , ... in a high-pass filter (59) for obtaining filter samples v (k), v (k + 1), v (k + 2), ... filtered and structure-borne noise occurring by estimating a statistical moment of at least second order, E {V ² } (k), E {V ² } (k + 1), E {V ² } (k + 2) of the filter samples v (k), v (k + 1), v (k + 2), ... is determined, where k is an integer as a run variable for the elapsed time.

2. The method according to claim 1, wherein the rotational speed is determined as the operating parameter, comprising the following steps: a) determining a starting value for the estimated value E (x ^* (k = 0)) as the expected value of the signal (40) for a starting time k = 0 and sampling the signal (40) at the time k = 0 to determine its instantaneous value as the sample start value x ^* (0) and then increasing the run variable k by 1, b) sampling the signal (40) at the time k + 1 to determine it Current value as (k + l) th sample x ^* (k + 1), c) calculating (S100) a (k + 1) th estimate E (x ^* (k + 1)) as the sum of the kth estimate E (x * (k)) and the product of an adaptation quantity cO with the result of subtracting the kth estimated value E (x * (k)) from the (k + 1) th sample value x ^* (k + 1), d) subtraction (S101) of the (k + 1) -th estimated value E (x ^* (k + 1)) calculated in step c) from the k + 1-th sample value x * (k + 1) determined in step b) and using the difference as (k + 1) th computation value tx (k + 1), e) comparison (S102) of the sign of the (k + 1) th calculated value x (k + 1) calculated in step d) with the sign of the corresponding kth calculated value x (k) from the previous one Running through step d), f) if the signs compared in step e) are the same: increase k by 1 and repeat steps b) to f); otherwise, comparison (S203) as to whether the (k + 1) th calculation value x (k + 1) of the current step d) is greater than the corresponding kth calculation value x (k) from the previous step of step d), g) if in step f) x (k + 1)> x (k) results: determination (S210 , S212, S214) of the point in time (t _on i. T _on2 ) at which the sign between the two calculated values determined in step e) changed with unequal signs in this ascending direction, and k increased by 1 and the steps were repeated b) to g) until x (k + 1)> x (k) has again resulted in step f); otherwise, determining (S230, S232, S234) the point in time (t _ab1 , t _ab2 ) at which the sign between the two calculated values determined in step e) changed with unequal signs in this falling direction, and increasing k by 1 and Repeat steps b) to g) until step f) has again shown that x (k + 1)> x (k) does not apply, and h) formation (S404; S504) of the difference (ΔT _aUf ; ΔT _a ) between the times determined in the different runs in step g) (t _aut ,, t _auf2 ; t _ab1 , t _ab2 ) and output (S510; S410) of this difference (ΔT _auf ; ΔT _a ) as a value that corresponds to the rotational speed of the roller bearing (20), and preferably continues the method from step b), where the following applies to cO: 0 <cO <1.

3. The method according to claim 2, characterized in that the times specified in the different runs in step g) (t _auf1,; t _ab ι, t _{a 2} ) by interpolation (S212; S232) between the two calculated values with an unequal sign Step f) are determined.

4. The method according to any one of the preceding claims, wherein the axial force is determined as the operating parameter and wherein the sensor arrangement (29, 36, 37, 38, 39) is designed such that the wave-shaped signal (40) when an axial force is applied to the roller bearing ( 20) changes, comprising the following steps: a) definition of a starting value for the estimated value E (x ^* (k = 0)) as the expected value of the signal (40) for a starting time k = 0 and sampling of the signal (40) at the time k = 0 to determine its instantaneous value as the sample start value x * (0) and then increase the run variable k by 1, b) sampling the signal (40) at the point in time k + 1 to determine its instantaneous value as the (k + 1) th sample value x ^* (k + 1), c) computing (S100) a (k + 1) th estimate E (x * (k + 1)) as the sum of the k th estimate E (x * (k)) and the product an adaptation variable cO with the result of subtracting the kth estimated value E (x ^* (k)) from the (k + 1) th sample value x ^* (k + 1), d1) Subt Reaction (S101) of the (k + 1) -th estimated value E (x * (k + 1)) calculated in step c) from the k + 1-th sample value x ^* (k + 1) determined in step b) and addition an auxiliary variable (stroke hlf) and use of the result as (k + 1) th calculated value x (k + 1), e1) comparison (S102) of the sign of the (k + 1) th calculated value x (k + 1) calculated in step d1) with the sign of the corresponding kth calculated value x (k) from the previous passage through step d1), f1 ) if the signs compared in step e1) are the same: increase k by 1 and repeat steps b) to f1); otherwise, comparison (S203) as to whether the (k + 1) th calculation value x (k + 1) from the current step d1) is greater than the corresponding kth calculation value x (k) from the previous execution of step d1), g1) if in step f1) there is x (k + 1)> x (k): definition (S210, S212, S214) of the point in time ( _starts , t _open2 ) at which the sign between the two was determined in step e1) Changed calculation values with unequal signs in this ascending direction, and increasing k by 1 and repeating steps b) to g1) until x (k + 1)> x (k) again resulted in step f1); otherwise, determining (S230, S232, S234) the point in time (t _ab1 , t _ab2 ) at which the sign between the two calculated values determined in step e1) changed with unequal signs in this falling direction, and increasing k by 1 and Repeat steps b) to g1) until step f1) has again shown that x (k + 1)> x (k) does not apply. i1) subtracting (S402) the point in time (t _ab2 ) specified in step g1) in the second alternative from the point in time (t ^) defined in step g1) in the first alternative and using the difference as the time interval (Δt_nhw) of the negative half-wave, d2) subtraction (S601) of the (k + 1) -th estimated value E (x * (k + 1)) calculated in step c) and the auxiliary variable (stroke hlf) from the k + 1-th sample value determined in step b) x * (k + 1) and use of the result as (k + 1) th calculation value y (k + 1), e2) comparison (S602) of the sign of the (k + 1) th calculation value y calculated in step d2) (k + 1) with the sign of the corresponding kth calculation value y (k) from the previous execution of step d2), f2) if the signs compared in step e2) are the same: increase k by 1 and repetition of steps b) to f2); otherwise, comparison (S603) as to whether the (k + 1) th calculation value y (k + 1) from the current step d2) is greater than the corresponding kth calculation value y (k) from the previous execution of step d2), g2) if in step f2) results y (k + 1)> y (k): definition (S610, S612, S614) of the point in time (t ^* _aut ι, t * _auf2 ) at which the sign between the two in step e2) changed the calculated values with unequal signs in this ascending direction, and increasing k by 1 and repeating steps b) to g2) until in step f2) again y (k + 1)> y (k) has resulted; otherwise determination (S630, S632, S634) of the point in time (t ^* _{ab1 l} t ^* _{a 2} ) at which the sign between the two calculated values determined in step e2) has changed with an unequal sign in this falling direction, and an increase in k by 1 and repetition of steps b) to g2) until it has again emerged in step f2) that y (k + 1)> y (k) does not apply, i2) subtraction (S638) of the step g2) in the first alternative time (t * _aUf2 ) from the time in step g2) in the second alternative (t ^* _ab2 ) and using the difference as the time interval (Δt_phw) of the positive half-wave, wherein steps d2) to i2) preferably run as far as possible parallel to steps d1) to i1), and j) determining (S408; S508) the difference (Δt_phw - Δt_nhw) between those in steps i1) and i2) Time intervals (Δt_nhw, Δt_phw) and output (S510; S410) of the sign (ax_dir) of this difference (Δt_phw - Δt_nhw) as a value which corresponds to the direction of the axial force acting on the roller bearing (20), and preferably continuing the method from step b ), where for cO applies: 0 <cO <1.

5. The method according to claim 4, characterized in that the times specified in steps g1) and g2) (t _auf1 , _∑ ; t _ab2 ; t ^* _auf1l t * _auf2 ; t * _ab1 , t _b2 ) by interpolation (S212; S232, S612; S632) can be determined between the respective two calculated values with an unequal sign from steps f1) or f2).

6. The method according to claim 4 or 5, characterized in that a calculation (S409, S509) of the amount (ax_for) of the difference formed in step j) is carried out and output as a measure of the magnitude of the axial force.

7. The method according to claim 1, wherein the radial force is determined as the operating parameter, comprising the following steps: a) establishing a starting value for the estimated value E (x * (k = 0)) as the expected value of the signal (40) for a starting time k = 0 and sampling the signal (40) at time k = 0 to determine its instantaneous value as sampling start value x ^* (0) and then increasing the running variable k by 1, b) sampling the signal (40) at time k + 1 to determine from its instantaneous value as (k + 1) th sample x ^* (k + 1), c) calculating (S100) a (k + 1) th estimate E (x * (k + 1)) as the sum of the k th estimate E (x * (k)) and the product of an adaptation quantity cO with the result of subtracting the k-th estimate E (x ^* (k)) from the (k + 1) th sample x * (k + 1 ), d) subtracting (S101) the (k + 1) -th estimated value E (x * (k + 1)) calculated in step c) from the k + 1-th sample value x * (k + 1) and use the difference as (k + 1) th calculation value x (k + 1), e) comparison (S102) of the sign of the (k + 1) th calculation value x (k + 1) calculated in step d) with the sign of the corresponding k th calculated value x (k) from the previous execution of step d), k) if the signs compared in step e) are the same: check (S302) whether the sign of the (k + 1) th calculated value x (k + 1) of step d) is positive, and if there is a positive sign: repetition of steps b) to k) until an unequal sign is detected in step e) and storage (S305) of the maximum (max_value) of the previously in step d ) calculated (k + 1) th calculation values x (k + 1), otherwise repeating steps b) to k) until an unequal sign is detected in step e), and storing (S315) the minimum (min_value) of the previously in step d) calculated (k + 1) th calculated values x (k + 1) and calculation of the difference (stroke) between the maximum (max_value) and the minimum (min_value) of these calculated (k + 1) th calculated values x (k + 1) and output (S108) of this difference (stroke) as a value which corresponds to the radial force acting on the roller bearing (20), and preferably increase k by 1 and continue the method from step b); if the signs compared in step e) are not the same: resetting the maximum (max_value) or the minimum (min_value) of the (k + 1) calculated values x (k + 1) calculated in step d) to 0 and increasing k by 1 and continuation of the method with step b), where for cO: 0 <cO <1.

8. The method according to claim 7, characterized in that in step k) for a positive sign, a check (S304) is carried out to determine whether the (k + 1) th calculation value x (k + 1) from step d) is greater than that previously determined maximum (max_value), and if so: storage (S305) of this (k + 1) th calculation value x (k + 1) as a new maximum (max_value) for the calculation (S308) of the difference (stroke) , and for a negative sign, a check (S314) is carried out to determine whether the (k + 1) th calculation value x (k + 1) from step d) is less than the previously determined minimum (min_value), and if so : Storage (S315) of this (k + 1) th calculation value x (k + 1) as a new minimum (min_value) for the calculation (S308) of the difference (stroke).

9. The method according to claim 7 or 8, characterized in that the first time through step k) the difference (hub) is additionally stored as a maximum difference (hub_max) and each time step k) is performed, a comparison (S110) is carried out whether the value of the currently calculated difference (hub) is greater than the maximum difference (hub_max), and if so: storage of this currently calculated difference (hub) as a new maximum difference (hub_max), which can preferably be deleted by a user.

10. The method according to any one of the preceding claims, wherein the direction of rotation is determined as the operating parameter and the sensor arrangement (27, 31, 32 ', 33, 34') is designed such that the wave-shaped signal (40) is different with different directions of rotation, comprising the following steps: a) definition of a starting value for the estimated value E (x ^* (k = 0)) as the expected value of the signal (40) for a starting point in time k = 0 and sampling of the signal (40) at the point in time k = 0 to determine it instantaneous value x as Abtaststartwert ^* (0), and then increasing the control variable k to 1, b) sampling the signal (40) at time k + 1 x for the determination of the instantaneous value as a (k + 1) -th sample ^* (k + 1 ) c) Calculating (S100) a (k + 1) -th estimated value E (x ^* (k + 1)) as the sum of the k-th estimated value E (x ^* (k)) and the product of an adaptation variable cO with the result of a Subtracting the kth estimate E (x ^* (k)) from the (k + 1) th sample x * (k + 1), d) Subtracting (S101) the (k + 1) calculated in step c) - th estimated value E (x * (k + 1)) from the k + 1-th sample value x ^* (k + 1) determined in step b) and using the difference as the (k + 1) -th calculated value x (k + 1 ), e) Comparison (S102) of the sign of the (k + 1) th calculated value x (k + 1) calculated in step d) with the sign of the corresponding k th calculated value x (k) from the previous passage through step d) , I) if the signs compared in step e) are the same: check (S302) whether the sign of the (k + 1) th calculation value x (k + 1) of step d) is positive, and if there is a positive sign results in: repetition of steps b) to I) until an unequal sign is detected in step e) and storage (S306) of the time Point (max_wert_t) of the occurrence of the maximum of the (k + 1) th calculated values x (k + 1) previously calculated in step d), otherwise repeating steps b) to I) until an unequal sign is detected in step e) , and storing (S316) the point in time (min_wert_t) of the occurrence of the minimum of the (k + 1) th calculated values x (k + 1) previously calculated in step d), as well as increasing k by 1 and continuing the method from step b ); if the signs compared in step e) are not the same: m) evaluation (S203; S204, S206; S224, S226) of the temporal reference of the point in time (max_wert_t) of the occurrence of the maximum of the previously calculated (k + 1) in step d) - th calculation values x (k + 1) and the time (min_wert_t) of the occurrence of the minimum of the (k + 1) th calculation values x (k + 1) previously calculated in step d) relative to one another for determining the direction of rotation (sym) of the rolling bearing (20) and its output (S208; S228), as well as increasing k by 1 and continuing the process with step b), where for cO: 0 <cO <1.

11. The method according to claim 10, characterized in that step m) contains a comparison (S203) as to whether the (k + 1) th calculation value x (k + 1) from the current step d) is greater than the corresponding kth calculation value x (k) from the previous step of step d), and includes the following step: n) if the comparison in step m) is positive: subtraction (S204) of a minimum auxiliary variable (min_t_ref) from the time (min_wert_t) of the occurrence of the minimum of (k + 1) th calculated values x (k + 1) previously calculated in step d), division of the result by 2, addition of the minimum auxiliary variables (min_t_ref) and subtraction of the point in time (max_wert_t) of the occurrence of the maximum of the previously in step d ) calculated (k + 1) th calculated values x (k + 1) thereof and subsequent determination of the sign (sym) of the result, then acceptance (S206) of the point in time (min_wert_t) of the occurrence of the minimum of the previously calculated in step d) ( k + 1) th calculation values x (k + 1) as new M minimum auxiliary variable (min_t_ref) and output (S208) of this sign (sym) as a value which corresponds to the direction of rotation of the rolling bearing (20), and increase of k by 1 and continuation of the method with step b); otherwise: subtraction (S224) of a maximum auxiliary variable (max_t_ref) from the time (max_wert_t) of the occurrence of the maximum of the (k + 1) th calculated values x (k + 1) previously calculated, division of the result by 2, addition the maximum auxiliary variable (max_t_ref) for this and subtraction of the point in time (min_wert_t) of the occurrence of the minimum of the (k + 1) calculated values x (k + 1) previously calculated in step d) thereof and subsequent determination of the sign (sym) of the result, then acceptance (S226) of the time (max_wert_t) of the occurrence of the maximum of the (k + 1) th calculated values x (k + 1) previously calculated in step d) as a new maximum auxiliary variable (max_t_ref) and output (S228) of this sign (sym ) as a value that corresponds to the direction of rotation of the roller bearing (20), as well as increasing k by 1 and continuing the process from step b),

12. The method according to claim 10 or 11, characterized in that in step I) for a positive sign, a check (S304) is carried out to determine whether the (k + 1) th calculation value x (k + 1) from step d) is greater than the previously determined maximum (max_wert), and if so: storage (S306) of the time (max_wert_t) of the occurrence of this (k + 1) th calculation value x (k + 1), and a check for a negative sign (S314) is carried out whether the (k + 1) th calculation value x (k + 1) from step d) is smaller than the previously determined minimum (min_value), and if so: storage (S316) of the time ( min_wert_t) of the occurrence of the minimum of this (k + 1) th calculation value x (k + 1).

13. The method according to any one of the preceding claims, wherein structure-borne noise events that act on the rotating roller bearing (20) are determined as the operating parameters, comprising the following steps: o) sampling the signal (40) at the time k = 0 to determine its instantaneous value as Initial sample value x ^* (0), p) filtering (S702) the initial sample value x * (0) obtained in step o) in a high-pass filter to obtain a filter sample value v (k = 0), q) definition of the second ordinary statistical moment E {V ² } (k = 0) of the filter sample value v (k = 0) as the starting torque value and then increasing the run variable k by 1, b) sampling the signal (40) at the time point k + 1 to determine its instantaneous value as (k + 1) -th sample x ^* (k + 1), r) filtering (S702) the sample x * (k + 1) obtained in step b) in a high-pass filter to obtain a filter sample v (k + 1), s) calculation (S704 ) of the second ordinary statistical moment u (k) = E {V ² } (k + 1) to the fil terabtastwert v (k + 1) as the sum of the second ordinary statistical moment E {V ² } (k) to the filter sampling value v (k) and the value of the result ddeerr SSuubbttrraakkttiioonn ddiieesseess multiplied by an adaptation quantity ddiieesseess two second ordinary statistical moment E {V ² } (k) of the square v ² (k + 1) of the filter sample v (k + 1), t) Output (S720) of a corresponding qualitative error signal or output (S710; S716) of a corresponding quantitative error value (output; Output_max) and preferably increase k by 1 and continue the method from step b), where d applies: 0 <d <1.

14. The method according to claim 13, characterized in that the following steps are carried out between steps s) and t): u) calculation (S706) of the second ordinary statistical moment E {U ² } (k + 1) to u (k + 1) as the sum of the second-order expected value E {U ² } (k) of u (k) and the value of the result of the subtraction of the second ordinary statistical moment E {U ² } (k) multiplied by an adaptation quantity c2 from the square u ² (k + 1) of the second ordinary statistical moment u (k + 1), v) calculation (S707) of the expected value E {U} (k + 1) of u (k + 1) as the sum of the expected value E {U} ( k) of u (k) and the value of the result of subtracting the expected value E {U} (k) from the second ordinary statistical moment u (k + 1) multiplied by an adaptation variable c3, w) calculating (S708) the variance σ ² (k + 1) as the value of the result of the subtraction of the square of the result of step v) from the result of step u), where for c2 and c3: 0 <c2 <1, 0 <c3 <1.

15. The method according to claim 14, characterized in that step s) is skipped and the filter sample v (k) is used instead of the second ordinary statistical moment u (k).

16. The method according to any one of claims 13 to 15, characterized in that the maximum of the quantitative error values (output) from step t) is stored (S714) as the maximum error value (Output_max).

17. The method according to any one of claims 2 to 16, characterized in that one or optionally more of the adaptation variables cO, d, c2 or c3 selected as a constant to minimize the calculation effort or to maximize the calculation accuracy, the speed and / or the calculation stability as dynamic variable can be selected.

18. The method according to any one of claims 4 to 17, characterized in that when determining a plurality of operating parameters, such steps, which are labeled with the same letters, do not need to be repeated.

19. Computer program product for determining the rotational speed of a roller bearing (20), which comprises a sensor arrangement (26; 27; 29) attached to the roller bearing (20), which has a wave-shaped signal (40) which is dependent on its rotational position when the roller bearing (20) rotates. provides, characterized in that it contains means for performing the steps according to at least one of the methods according to claims 1 to 18.

20. Rolling bearing with z rolling elements (24), where z is an integer, preferably an even number, and with a device for determining operating parameters of the rolling bearing (20), comprising at least one, possibly more than z and preferably up to z / 2 sensor arrangements (26; 27; 29) which are distributed in the circumferential direction of the rolling bearing (20) and are attached to the rolling bearing (20), each sensor arrangement (26; 27; 29) generating a wave-shaped signal which depends on its rotational position when the rolling bearing (20) rotates (40) supplies, and at least one, possibly more than z and preferably up to z / 2, evaluation devices (50), each evaluation device (50) being connected to a sensor arrangement (26; 27; 29), each evaluation device ( 50) contains an evaluation unit (51-55) for executing the steps according to at least one of the methods according to claims 1 to 18.

21. Rolling bearing according to claim 20, characterized in that each sensor arrangement (26; 27; 29) is formed from four strain gauges (31-34; 31, 32 ', 33, 34'; 36-39) connected to form a Wheatstone bridge , which are arranged one behind the other in the circumferential direction of the roller bearing (20), the first strain gauge (31; 36) and the third strain gauge (33; 38) of each sensor arrangement (26; 27; 29) and the second strain gauge (32; 32 '; 37) and the fourth strain gauge (34; 34 '; 39) of each sensor arrangement (26; 27; 29) in the circumferential direction have n times the distance between two adjacent rolling elements (24), or two strain gauges connected to a Wheatstone half bridge is formed, which are arranged one behind the other in the circumferential direction of the roller bearing (20), the signal (40) being generated by rolling the strain gauges (31-34; 31, 32 ', 33, 34') through the roller bodies (24) and n is an integer he or equal to 1.

22. Rolling bearing according to claim 21, characterized in that the strain gauges (31-34; 31, 32 ', 33, 34'; 36-39) and the at least one, preferably up to z 2, evaluation devices (50) in a rotating Outer groove (23) of the outer ring (22) of the rolling bearing (20) are arranged.

23. Rolling bearing according to one of claims 20 to 22, characterized in that each evaluation unit (51-55) and preferably each evaluation device (50) is designed as an electrical circuit, preferably as an ASIC.